Knowledge Graph Embedding Prediction for New Node Links
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current knowledge graph embedding methods require re-training using the entire dataset to predict authenticity for new data, resulting in high prediction costs.
Innovation Solution
An information processing device that determines whether predicted graph data includes a node link not in the training data, specifies similar graph data from training data based on embedding predictions, and calculates prediction results using these similar data points, thereby reducing the need for re-training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If re-training using the entire dataset is performed to predict authenticity for new data, then prediction accuracy is improved, but prediction cost increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating embedding vectors for all training data before prediction. The system embeds the entire training dataset in advance, creating a ready-to-use vector representation that can be quickly queried during prediction without requiring re-training. This pre-computation stores the essential knowledge structure in vector form, enabling fast similarity-based predictions for new data points.
Solution Approach 2:
The patent uses copying by creating embedding vectors as simplified representations of the original graph data. Instead of working with the complex original graph structure during prediction, the system copies the essential relational information into vector embeddings. These vector copies enable efficient distance-based similarity calculations while preserving the semantic relationships needed for accurate prediction.
2Speed
If embedding vectors are calculated for all training data in advance, then prediction speed is improved, but memory usage increases
Solution Approach 1:
The patent extracts only the essential embedding vectors needed for prediction from the complete training dataset. Rather than storing all possible graph configurations and their relationships, the system extracts and stores only the condensed vector representations that capture the fundamental relational patterns. This extraction reduces the stored data to its most compact useful form, balancing memory efficiency with prediction capability.
3Loss of energy
If similarity-based prediction using embedding vectors is used, then prediction cost is reduced, but handling of completely new node links may be less accurate
Solution Approach 1:
The patent applies parameter changes by measuring similarity through distance metrics in the embedding vector space. Instead of requiring exact matches or re-training on new data, the system changes the approach to prediction by calculating distances between embedding vectors. This parameter-based similarity measurement allows the system to handle new node links by finding the closest existing patterns in the training data, providing reliable predictions without re-training while keeping costs low.
Data Source
AI summary
A non-transitory computer-readable recording medium storing a prediction program that uses knowledge graph embedding, for causing a computer to execute processing including: determining whether or not graph data to be predicted is data that includes a node link that indicates a relationship between nodes not included in training data used for training of the knowledge graph embedding; specifying, in a case where it is determined that the graph data to be predicted is the data that includes the node link not included in the training data, graph data similar to the graph data to be predicted from the training data based on a result of embedding prediction for a label of a node included in the graph data to be predicted; and determining a prediction result for the graph data to be predicted based on the specified similar graph data.